PhysMoE: Fast Spatial Grounding with Single-Token Coordinate Representation in Modality-Decoupled VLMs
cs.CV, cs.AI, cs.RO
Submitted: 2026-11-08
Updated: 2026-11-08
Terminology
Sources
- Gemini Robotics: Bringing AI into the Physical World
- GPT-4 Technical Report
- Cosmos 3: Omnimodal World Models for Physical AI
- Cosmos-Reason1: From Physical Common Sense To Embodied Reasoning
- Qwen3-VL Technical Report
- Shikra: Unleashing Multimodal LLM's Referential Dialogue Magic
- RoboAfford++: A Generative AI-Enhanced Dataset for Multimodal Affordance Learning in Robotic Manipulation and Navigation
- Mixtral of Experts
- RynnBrain 1.1: Towards More Capable and Generalizable Embodied Foundation Model
- RoboBrain 2.5: Depth in Sight, Time in Mind
- LLaMA: Open and Efficient Foundation Language Models
- Point-It-Out: Benchmarking Embodied Reasoning for Vision Language Models in Multi-Stage Visual Grounding
- ERQA-Plus: A Diagnostic Benchmark for Reasoning in Embodied AI
- Qwen3-VL-Seg: Unlocking Open-World Referring Segmentation with Vision-Language Grounding
- Embodied-R1.5: Evolving Physical Intelligence via Embodied Foundation Models
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs